2/21/2026
Show HN: Murl – Curl for MCP Servers
Command-line tool for interacting with MCP servers, simplifying JSON-RPC interactions similar to curl for developers.
2/21/2026
Show HN: AI Council – multi-model deliberation that runs in the browser
Browser-based tool for multi-model LLM deliberation supporting local (Ollama) and cloud models with zero server/storage requirements.
2/21/2026
Show HN: Natural language search across Kalshi and Polymarket (API and MCP)
Natural language search API and MCP interface for querying 80k prediction market contracts, enabling AI agents to access market data.
2/20/2026
The enviromental impact of using LLMs for writing code
Analysis of environmental impact and energy consumption from using LLMs for code generation and development tasks.
2/20/2026
AI Agents Now Consume More Tokens Than Humans
Analysis of LLM token consumption patterns showing AI agents now consume more tokens than human users.
2/20/2026
Show HN: Local AI document intelligence – no cloud, runs on your machine
Local document intelligence platform with LangGraph multi-agent pipeline, OCR, and semantic search running entirely offline on user machines.
2/20/2026
Aslan Browser: Open-sourced a macOS browser for AI agents
Open-source native macOS browser for AI agents using WKWebView and Unix sockets, designed for agent automation without full Chrome overhead.
2/20/2026
Apple researchers develop on-device AI agent that interacts with apps
Apple research on Ferret-UI Lite, a 3B-parameter on-device multimodal agent that understands UI elements and user interactions.
2/20/2026
I Sold Out for $20 a Month and All I Got Was This Perfectly Generated Terraform
Personal experience using LLM code generation tools (Copilot, Gemini) for infrastructure automation, evaluating practical utility.
2/20/2026
Show HN: OpenClaw Live2D – Open-source AI companion with Live2D avatar
Open-source frontend framework for AI companions with Live2D avatars, long-term memory, and emotional affinity systems.
2/20/2026
Theia: Hold Your LLM Accountable
Chrome extension for fact-checking claims and detecting hallucinations in web content with source verification.
2/20/2026
What If LLMs Could See Trust Boundaries?
Technical analysis of prompt injection attacks in agentic systems and proposal for LLMs to recognize trust boundaries in untrusted content.
2/20/2026
AI Influencer Playbook
Case study on building and scaling AI-generated video influencer accounts on TikTok, documenting strategies and metrics.
2/20/2026
HalluciGuard – open-source middleware to detect and mitigate LLM hallucinations
Open-source middleware library that detects and mitigates hallucinations in production LLM applications with a reliability wrapper layer.
2/20/2026
Detecting Elements of Claude's Architectural Image
Analysis exploring Claude model architectural elements through self-reflection mechanisms and public model interrogation.
2/20/2026
Turn Your LLM into a Calibrated Classifier for $2
Method to convert LLMs into calibrated classifiers with minimal cost.
2/20/2026
Your AI Agent Will Make Money. Here's How It Traces Back to You.
Essay exploring accountability and authorization tracking systems needed as AI agents perform autonomous financial and decision-making tasks.
2/20/2026
Show HN: MMORPG for AI agents – Openbattle.club
OpenBattle.club is a competitive Pokemon-themed MMORPG environment for testing and training AI agents with live battles and leaderboards.
2/20/2026
Perfect agreement is a warning sign you're talking to yourself
Analysis of AI agent failure modes when perfectly aligned with outdated assumptions, highlighting debugging challenges.
2/20/2026
Show HN: Vipune – Simple Memory for Agents
Local CLI memory store for AI agents using SQLite embeddings with semantic search and recency reranking.
2/20/2026
Show HN: Created an open-source QA agent
Open-source QA agent built with agent-browser, ai-sdk, and llmgateway.
2/20/2026
Show HN: Delegate – AI eng manager running team of agents for async multitasking
Browser-based tool managing teams of AI coding agents with peer review, task breakdown, and git management for async development.
2/20/2026
Multi-Turn Intent Detection for LLM and Agent Security (ArXiv)
ArXiv research on detecting malicious intents across multi-turn conversations in LLM and agent systems to improve security.
2/20/2026
Show HN: Cerebro – I got tired of waiting for Apple to fix their keyboard
Custom iOS keyboard leveraging AI for text rewriting and language switching.
2/20/2026
Show HN: AgentCheck – AI bot posture leaderboard from robots.txt and llms.txt
Open-source leaderboard tracking AI bot policies via robots.txt and llms.txt declarations, showing which sites allow/block agents.
2/20/2026
AI coding assistant Cline compromised to create more OpenClaw chaos
Security breach in Cline AI coding assistant related to OpenClaw vulnerability.
2/20/2026
What do you do after paste in sensi info i.e: API token & key into IDE AI chat?
Discussion about security practices when pasting sensitive credentials into IDE AI chat interfaces.
2/20/2026
Mixing the new GitHub agent workflow and Jira MCP to accelerate PR review
Example combining GitHub agent workflow with Jira MCP for accelerated pull request review.
2/20/2026
Taalas Etches AI Models onto Transistors to Rocket Boost Inference
Hardware approach using SRAM and tensor engines to accelerate AI inference compared to GPU alternatives.
2/20/2026
A chat room where LLM bots pretend to be human and everyone hunts each other
Interactive chat application where LLM bots pose as humans in competitive gameplay scenario.
2/20/2026
Prompt Repetition Improves Non-Reasoning LLMs
Research showing that repeating prompts multiple times improves accuracy in non-reasoning LLM tasks.
2/20/2026
Show HN: Ember MCP – local persistent memory for LLMs, kills stale memories
Open-source memory system for LLMs using shadow-decay mechanism to mark stale memories, integrates via MCP protocol with vector storage.
2/20/2026
I built a live honeypot that catches AI agents. Here's what happened
Honeypot experiment capturing AI agent behavior and security vulnerabilities in autonomous systems.
2/20/2026
If LLMs Only Predict the Next Token, Why Do They Work?
Technical analysis of why next-token prediction in LLMs produces emergent capabilities despite simple training objective.
2/20/2026
Amazon blames human employees for an AI coding agent's mistake
News story about Amazon's AI coding agent making mistakes attributed to human oversight failures.
2/20/2026
Raison – Version control and real-time deployment for AI prompts
Raison: version control and real-time deployment platform for AI prompts with management capabilities.
2/20/2026
Intuitive Intro to Reinforcement Learning for LLMs
Educational content explaining reinforcement learning techniques applied to LLM training and optimization.
2/20/2026
I built an agent that reads Jira tickets and opens pull requests automatically
AI agent that automatically reads Jira tickets and creates pull requests, demonstrating autonomous workflow automation.
2/20/2026
Skillflag: CLI flag convention for listing and installing agent skills
Chess learning app integrating Claude AI coach with Stockfish engine via MCP for real-time move evaluation.
2/20/2026
Show HN: BeadHub, Beads-based coordination for multiple coding agents
BeadHub: open-source coordination layer for multiple AI agents using Beads git-native tracking with agent-to-agent chat.
2/20/2026
Fall-from-grace: A prompt engineering functional programming language
CLI flag convention standard for listing and installing agent skills in AI agent systems.
2/20/2026
The AI security nightmare is here and it looks suspiciously like lobster
Security vulnerability: hacker exploited Claude-powered Cline agent workflow to install malware OpenClaw.
2/20/2026
I hacked ChatGPT and Google's AI – and it only took 20 minutes
Demonstration of prompt injection attacks on ChatGPT and Google AI causing false outputs about hot dogs.
2/20/2026
The splines are hallucinating now: how I built and what got built by AI mayors
Hallucinating Splines: platform where AI agents play Micropolis city simulation via REST API, built with Claude.
2/20/2026
Show HN: Skills – Making AI coding tools aware of government standards
Skills: Markdown-based domain knowledge injection for AI coding tools to enforce government standards and protocols.
2/20/2026
Show HN: LLMWise – Compare, Blend, and Judge LLM Outputs from One API
LLMWise API provides orchestration for combining multiple LLMs with Mixture-of-Agents blending and comparison features.
2/20/2026
Warden
AI code review agent tool using Claude with skill-based architecture for local CLI and GitHub Actions integration.
2/20/2026
Show HN: An offline-first ski analysis app
Offline-first iOS ski technique analysis app built by solo developer without cloud backend or subscriptions.
2/20/2026
Parallel AI Agents That Review My Code (Claude Code Setup)
Using Claude as parallel AI agents to review code automatically in development workflow.
2/20/2026
Raising Agentic Children
Discussion on challenges of AI agents - their limitations, unpredictability, and failure modes when deployed.